1. Field of the Invention
The present invention relates to voice access to data systems in general, and a method and system for updating grammars that are used in voice access systems in particular
2. Background Information
Recently, the development of new technologies in the telephony and speech recognition areas has enabled data systems to be accessed. via voice (e.g., spoken verbal input). Such voice-enabled data systems let users navigate to request a particular piece of data or a related set of data using voice commands, and then “reads back” the data to the user. These types of data systems are widely used in information systems in various industries, including banking, insurance, investments, etc.
In general, .in order to provide voice access to a data system, a complete voice interface must be provided. This typically involves integrating various hardware and software components, such as telephony interfaces, multi-channel access equipment, speech recognition software, text-to-speech (TTS) software and hardware, and other related components. A fundamental aspect of the a voice access system concerns the grammar used by the system. A grammar is a specification of all allowed user utterances (syntax) in an application, such as those used a voice application, which contains the logic for enabling interactions between a voice interface and a data system. A grammar also contains the meanings of each key word or phrase (semantics) corresponding to the user utterances.
Typically, a grammar will include information pertaining to specific data in the data system that is accessed via the voice interface. For example, if a voice access system is to include access to contact information, the grammar will generally include specific contact information, such as last name, first name, address, etc. This information is used to disambiguate between different user utterances. For instance, suppose a desired contact name is Laura Smith, and the data system includes contacts named, Laura Smith, Lauren Smith, Lori Smith and Lorna Smith. Since each of these names sounds similar but are different, there needs to be specific grammar information for each name.
Another aspect of grammars concerns performance. In order to provide adequate performance, grammar information will generally be stored locally (i.e., on a server that is local to a voice application). In some implementations, the grammar is stored in a pre-compiled format for further performance enhancement. To support these architectures, it is necessary to first provide an initial set of grammar information pertaining to data in the data system, and then periodically update the grammar information to reflect changes in the data in the data system. Typically, this grammar update process will require a specific grammar update module or application to be written for each voice application.
Several vendors now provide integrated voice-access systems that handle the voice-to-computer aspects of a voice-enabled data system. For example, these integrated systems handle telephony connectivity to a phone network, perform speech recognition to convert voice commands and verbalized data requests to a computer-readable form (e.g., ASCII text or binary text form), and handle the TTS functions. Although these integrated voice-access systems provide a lot of the workload, they still require grammars to be written and updated to enable access to the data system.